• Gaussian blur sigma=2, convert to frequencies before pixel FlowSOM, normalize including calibration channel
  • Pixel hClusters found using z-score cap=3, k=20, and iterative thresh=2 (1 iteration)
  • Cell-based clustering using frequencing of 20 hClusters (cap): normalize by size
  • k=10 for cell clustering

1 Cell-based FlowSOM clustering using 100 pixel clusters

1.1 No calibration

1.1.1 Cell cluster x frequency of pixel clusters

1.1.2 Cell cluster x marker (not weighted, segmentation)

1.1.3 Cell cluster x marker (weighted by pixel cluster frequency)

1.1.4 Point breakdown

1.1.5 Compare frequency with calibration

1.2 Normalize using calibration

1.2.1 Cell cluster x frequency of pixel clusters

1.2.2 Cell cluster x marker (not weighted, segmentation)

1.2.3 Cell cluster x marker (weighted by pixel cluster frequency)

1.2.4 Point breakdown

1.2.5 Compare frequency with calibration

2 Cell-based FlowSOM clustering using hClusters (cap, k=20)

2.1 No calibration

2.1.1 Cell cluster x frequency of pixel clusters

2.1.2 Cell cluster x marker (not weighted, segmentation)

2.1.3 Cell cluster x marker (weighted by pixel cluster frequency)

2.1.4 Point breakdown

2.1.5 Compare frequency with calibration

2.2 Normalize using calibration

2.2.1 Cell cluster x frequency of pixel clusters

2.2.2 Cell cluster x marker (not weighted, segmentation)

2.2.3 Cell cluster x marker (weighted by pixel cluster frequency)

2.2.4 Point breakdown

2.2.5 Compare frequency with calibration